Nonuniformity correction for focal plane arrays and partition based superresolution of video with application to an infrared imaging system

The utility of many infrared imaging systems is known to be affected by two common problems. The images acquired using an infrared imaging system is corrupted by the spatial fixed pattern noise. Spatial fixed pattern noise is caused by the variation in the photoresponses of the individual detectors in the focal plane array and by the nonuniform response of the readout and digitization electronics involved in multiplexing the individual detector responses. In addition, the spatial nonuniformity tends to drift in time due to the variation in the surrounding conditions.; The second problem associated with any imaging system is that the detectors placed at the focal plane array of the imaging system are not close enough to sufficiently sample the scene at the Nyquist rate. The acquired images are degraded by the aliasing effects. The spatial frequencies could be sufficiently bandlimited for the chosen detector array by reducing the aperture of the optics. However, some aliasing is often preferred to a loss of optical resolution in the design of imaging systems.; In this dissertation, we describe a new scene based nonuniformity correction (NUC) algorithm that treats the aggregate spatial nonuniformity in separate stages. One advantage of the proposed algorithm is that it requires only few frames to obtain high quality corrections. The effectiveness of the proposed approach is tested by applying it to simulated and infrared imagery.; We also investigate the use of partition-based weighted sum (PWS) filters for super-resolution (SR) of video. Though, the PWS filters have been previously used for image super-resolution, they are not suitable for video processing. Moreover, some of the existing SR algorithms are computationally intensive and may be difficult to implement in real time video system. In this dissertation, we present a novel approach for applying the partition-based SR filters to video that is computationally efficient and suitable for parallel implementation. The previously proposed PWS filters were used for image SR with a translation motion model. Here, a new implementation of the PWS filters for a general motion model is proposed. We also present a detailed computational analysis, quantitative comparisons with several benchmark techniques, and apply and evaluate the method with rotational motion as well as translational motion.